hugodk-sch
commited on
Commit
•
fa7bcfd
1
Parent(s):
c02310c
End of training
Browse files- README.md +13 -11
- all_results.json +12 -12
- config.json +3 -1
- eval_results.json +13 -13
README.md
CHANGED
@@ -1,11 +1,13 @@
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
|
|
4 |
- trl
|
5 |
- dpo
|
6 |
-
- alignment-handbook
|
7 |
- generated_from_trainer
|
8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
9 |
model-index:
|
10 |
- name: aftonposten-6b-align-scan
|
11 |
results: []
|
@@ -16,17 +18,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
|
17 |
# aftonposten-6b-align-scan
|
18 |
|
19 |
-
This model is a fine-tuned version of [
|
20 |
It achieves the following results on the evaluation set:
|
21 |
-
- Loss: 0.
|
22 |
-
- Rewards/chosen: -0.
|
23 |
-
- Rewards/rejected: -0.
|
24 |
-
- Rewards/accuracies: 0.
|
25 |
-
- Rewards/margins: 0.
|
26 |
-
- Logps/rejected: -37.
|
27 |
-
- Logps/chosen: -34.
|
28 |
-
- Logits/rejected: -2.
|
29 |
-
- Logits/chosen: -2.
|
30 |
|
31 |
## Model description
|
32 |
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
4 |
+
- alignment-handbook
|
5 |
- trl
|
6 |
- dpo
|
|
|
7 |
- generated_from_trainer
|
8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
9 |
+
datasets:
|
10 |
+
- hugodk-sch/aftonposten_title_prefs
|
11 |
model-index:
|
12 |
- name: aftonposten-6b-align-scan
|
13 |
results: []
|
|
|
18 |
|
19 |
# aftonposten-6b-align-scan
|
20 |
|
21 |
+
This model is a fine-tuned version of [data/ap-gpt-j-6b-sft-qlora-04-08](https://huggingface.co/data/ap-gpt-j-6b-sft-qlora-04-08) on the hugodk-sch/aftonposten_title_prefs dataset.
|
22 |
It achieves the following results on the evaluation set:
|
23 |
+
- Loss: 0.9173
|
24 |
+
- Rewards/chosen: -0.0874
|
25 |
+
- Rewards/rejected: -0.1849
|
26 |
+
- Rewards/accuracies: 0.5743
|
27 |
+
- Rewards/margins: 0.0975
|
28 |
+
- Logps/rejected: -37.7220
|
29 |
+
- Logps/chosen: -34.1316
|
30 |
+
- Logits/rejected: -2.2003
|
31 |
+
- Logits/chosen: -2.2051
|
32 |
|
33 |
## Model description
|
34 |
|
all_results.json
CHANGED
@@ -1,18 +1,18 @@
|
|
1 |
{
|
2 |
"epoch": 4.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -34.
|
6 |
-
"eval_logps/rejected": -37.
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": -0.
|
10 |
-
"eval_rewards/margins": 0.
|
11 |
-
"eval_rewards/rejected": -0.
|
12 |
-
"eval_runtime": 145.
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
"train_loss": 0.15107146371881683,
|
17 |
"train_runtime": 10740.8737,
|
18 |
"train_samples": 3079,
|
|
|
1 |
{
|
2 |
"epoch": 4.0,
|
3 |
+
"eval_logits/chosen": -2.2051267623901367,
|
4 |
+
"eval_logits/rejected": -2.2002947330474854,
|
5 |
+
"eval_logps/chosen": -34.13163375854492,
|
6 |
+
"eval_logps/rejected": -37.722042083740234,
|
7 |
+
"eval_loss": 0.9173061847686768,
|
8 |
+
"eval_rewards/accuracies": 0.574335515499115,
|
9 |
+
"eval_rewards/chosen": -0.08737529814243317,
|
10 |
+
"eval_rewards/margins": 0.09750958532094955,
|
11 |
+
"eval_rewards/rejected": -0.18488489091396332,
|
12 |
+
"eval_runtime": 145.0469,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.365,
|
15 |
+
"eval_steps_per_second": 0.296,
|
16 |
"train_loss": 0.15107146371881683,
|
17 |
"train_runtime": 10740.8737,
|
18 |
"train_samples": 3079,
|
config.json
CHANGED
@@ -18,6 +18,8 @@
|
|
18 |
"n_layer": 28,
|
19 |
"n_positions": 2048,
|
20 |
"quantization_config": {
|
|
|
|
|
21 |
"bnb_4bit_compute_dtype": "float16",
|
22 |
"bnb_4bit_quant_type": "nf4",
|
23 |
"bnb_4bit_use_double_quant": false,
|
@@ -45,7 +47,7 @@
|
|
45 |
},
|
46 |
"tie_word_embeddings": false,
|
47 |
"tokenizer_class": "GPT2Tokenizer",
|
48 |
-
"transformers_version": "4.
|
49 |
"use_cache": true,
|
50 |
"vocab_size": 50400
|
51 |
}
|
|
|
18 |
"n_layer": 28,
|
19 |
"n_positions": 2048,
|
20 |
"quantization_config": {
|
21 |
+
"_load_in_4bit": true,
|
22 |
+
"_load_in_8bit": false,
|
23 |
"bnb_4bit_compute_dtype": "float16",
|
24 |
"bnb_4bit_quant_type": "nf4",
|
25 |
"bnb_4bit_use_double_quant": false,
|
|
|
47 |
},
|
48 |
"tie_word_embeddings": false,
|
49 |
"tokenizer_class": "GPT2Tokenizer",
|
50 |
+
"transformers_version": "4.39.0.dev0",
|
51 |
"use_cache": true,
|
52 |
"vocab_size": 50400
|
53 |
}
|
eval_results.json
CHANGED
@@ -1,16 +1,16 @@
|
|
1 |
{
|
2 |
-
"epoch":
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -34.
|
6 |
-
"eval_logps/rejected": -37.
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": -0.
|
10 |
-
"eval_rewards/margins": 0.
|
11 |
-
"eval_rewards/rejected": -0.
|
12 |
-
"eval_runtime": 145.
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
}
|
|
|
1 |
{
|
2 |
+
"epoch": 4.0,
|
3 |
+
"eval_logits/chosen": -2.2051267623901367,
|
4 |
+
"eval_logits/rejected": -2.2002947330474854,
|
5 |
+
"eval_logps/chosen": -34.13163375854492,
|
6 |
+
"eval_logps/rejected": -37.722042083740234,
|
7 |
+
"eval_loss": 0.9173061847686768,
|
8 |
+
"eval_rewards/accuracies": 0.574335515499115,
|
9 |
+
"eval_rewards/chosen": -0.08737529814243317,
|
10 |
+
"eval_rewards/margins": 0.09750958532094955,
|
11 |
+
"eval_rewards/rejected": -0.18488489091396332,
|
12 |
+
"eval_runtime": 145.0469,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.365,
|
15 |
+
"eval_steps_per_second": 0.296
|
16 |
}
|